Bad-Goods Analysis Platform
Overview
Our client, a Fortune Global 500 food and beverage leader, wanted a Bad-Goods Analysis Platform that could automate identification and escalation workflows, bring together live sales and returns data with smart thresholds and filters, deliver interactive dashboards with branch, business, and brand-level visibility, and give teams the user-friendly tools, automated alerts, and one-click exports.
Solution
- Developed the lead scoring model on a dataset of ~16.6 lakh leads with a 1.17% booking rate
- Conducted Exploratory Data Analysis (EDA) that established booking conversion varied significantly across source, geography, budget and lead timing behavior, and project characteristics
- Engineered features and built machine learning models using Logistic Regression, Random Forest, and XGBoost, iterating the framework from a single overall model to separate Direct, Walk-in, and CP models for more granular and accurate conversion prediction
- Implemented dynamic project-level Platinum tagging to maintain stable Platinum allocation across the portfolio and improve real-time usability of model outputs
Impacts Delivered
The final lead scoring model created a focused opportunity pool where a small subset of leads contributed disproportionately towards bookings:
- Delivered a fully operational Bad-Goods Analysis Platform that automates the identification and management of bad goods across the entire distribution network, bringing together live sales and returns data, smart thresholds, and interactive dashboards in a single, unified system.
- Established a scalable foundation for future automation and analytics initiatives.
- Processed 15+ million records with zero downtime, reducing reporting and coordination effort while resolving issues faster through live escalation workflows.
- Enhanced transparency and accuracy across brands and branches, giving a reliable, real-time view of bad-goods performance at every level of the distribution hierarchy.
